Google Analytics

Traffic Sources

Every site, engine and channel sending you sessions — the sites that link to you, search engines, social networks, AI answer engines and direct visits — with conversions for each. Search Console tells you who found you on Google; this tells you who sent you everyone else.

What it measures

  • Source — the hostname or engine credited with the session (pltechlabs.com, google, (direct)).
  • Medium — how GA4 categorised it (referral, organic, cpc, (none) for direct).
  • Channel — our grouping of the two, explained below.
  • Sessions and Share — visits from that source, and its percentage of all sessions in the window.
  • Conversions — GA4 key events credited to those sessions (0 if none are configured).

How the channel is decided

GA4's own Default Channel Group isn't available on the dimensions we ingest, so we derive the channel from the source and medium pair. It's plain rules, evaluated in this order — no AI, nothing to drift:

ChannelMatched when
AI answer engineSource contains chatgpt, openai, perplexity, gemini, claude or copilot
DirectSource is (direct), or medium is (none)
PaidMedium is cpc, ppc, display, ads or similar
SocialSource is a known social network
SearchMedium is organic, or the source is a search engine domain
ReferralMedium is referral — another site linked to you
UnknownSource is (not set), or nothing above matched

Order matters in two places. AI engines are checked first because several sit on a search-engine domain — gemini.google.com would otherwise read as Search. And Search is checked before Referral because Google labels some of its own surfaces search.google.com / referral; counting those as a third-party referral would overstate who is actually linking to you.

How to read it

  • Sort by conversions, not sessions. A site sending 40 visits that convert matters more than one sending 4,000 that bounce.
  • Referrals are the channel you can actually act on. Search and AI traffic is earned indirectly; a referring site is a real relationship — worth a thank-you, a reciprocal link, or a follow-up.
  • A new referrer appearing is worth investigating — someone wrote about you, and you probably don't know it yet.
  • The report calls out how many referring sites you have above the table, since they're easy to miss in a list dominated by direct and search.

Caveats

  • No users column, on purpose. Rows are warehoused per landing page, so once we collapse that dimension a visitor who landed on two pages from the same source would be counted twice. Sessions and conversions add up correctly; users don't, so we don't show a number we can't stand behind.
  • Direct is a bucket, not a channel. It collects genuine type-ins along with anything GA4 couldn't attribute — app clicks, some email, and traffic that lost its referrer over HTTPS.
  • Requires a connected GA4 property; the first 90-day import runs in the background after linking.
  • Conversions depend on GA4 key events being configured in your property.
  • Sources are as GA4 reports them. If your own domain appears, that's a self-referral — usually a cross-subdomain tagging gap rather than real traffic.

Related reports